To get an idea of how personalized medicine could reshape patient care in the years ahead, one need only look at how it is beginning to reshape the care of patients with cancer. Cancer is where personalized medicine has gained its firmest foothold. The longstanding scattershot practice of prescribing the same drugs to virtually all patients with a particular type of cancer is giving way to a more selective approach in which genetic tests are run on tumor samples to identify which patients are likely to benefit from which drugs, or to tell whether they need drugs at all. [End Page 18] This shift holds the prospect of improving cancer care and the stewardship of costly cancer care resources. It has had its successes and failures, its swells of enthusiasm and ebbs of skepticism. It has interjected the talk of genetic markers, like KRAS and EGFR, into the doctor-patient conversation, giving it the gloss of scientific certitude. But in reality much of the decision-making that is taking place now requires a giant leap of faith.
Nowhere are this shift and its mashup of effects more apparent than with breast cancer. Until just a few years ago, it was a given that most women with operable breast cancer—including many with early-stage disease—would get chemotherapy to prevent recurrence. Doctors and patients knew that chemo was a blunt instrument: it could increase survival but would also damage healthy cells, causing hair loss and other severe side effects. But doctors knew something else: that chemo would do more harm than good for many patients. Their risk of recurrence was so low that chemotherapy probably would not increase their survival. The trouble was, it was impossible to tell in advance who these women were. In the absence of a crystal ball, the safest decision seemed to be to give chemotherapy and hope for the best.
Now, there are a few crystal balls: genetic tests on the market that predict the likelihood of breast cancer recurring based on the expression of certain genes in a tumor sample. The test with the strongest evidence behind it, according to a panel sponsored by the Centers for Disease Control and Prevention to evaluate genetic tests, is Oncotype DX. It is designed for women newly diagnosed with early-stage breast cancer that is estrogen-receptor positive and has not spread to the lymph nodes. About half of the 207,000 women expected to get a breast cancer diagnosis this year will fall into this category. The test is cited in two breast cancer treatment protocols—from the American Society for Clinical Oncology and the National Comprehensive Cancer Network—as a tool for deciding whether to recommend chemotherapy, along with pathological features and other information that indicates a tumor's aggressiveness. Medicare and most private insurers cover the $3,900 cost of the test, and ninety thousand women have had it, fifty thousand of them in 2009 alone, according to Genomic Health, the manufacturer.
On its Web site, Genomic Health has revolving photographs of two chairs: a hospital chair next to an intravenous drip, and a wing chair next to an elegant table with fresh flowers and a mug. The words "Chemo? No Chemo?" are superimposed on the photographs. The implication is that Oncotype DX makes the choice simple and straightforward.
Does it? That depends. A look at how the test works reveals why the choice is often neither simple nor straightforward. The test measures the expression of twenty-one genes—sixteen related to cancer recurrence and disease-free survival and five noncancer genes included as reference points to verify the results. The results are used to calculate a "recurrence score" on a scale of one to one hundred. A score of one to seventeen indicates a low risk (about 7 percent) of distant recurrence within ten years, evidence that chemotherapy might not be beneficial. A score of thirty-one to one hundred indicates a high risk (about 30 percent) of recurrence, evidence that chemotherapy could be life-saving. But a score of eighteen to thirty indicates an intermediate risk (14 percent), and no evidence...